CN1357181A - Method and system for fast maximum posteriori decoding - Google Patents

Abstract

Processing delay associated with MAP decoding (and variants thereof) is reduced by processing forward and reverse state metrics in parallel. Further reductions in processing complexity can be achieved by omitting redundant operations and multiplexing processing units functionality.

Description

The method and system of fast maximum posteriori decoding

Background

The present invention relates generally to the fault processing in the field of wireless communications, more specifically to the signal that uses the decoding of maximum a posteriori probability (MAP) decoding technique with the error correcting code transmission.

The explosive increase of the development of business communication system, particularly cellular radiotelephone system forces the system designer seeks increases power system capacity under situation about communication quality not being reduced to outside the tolerable thresholding of user method.A kind of technology that reaches these targets relates to wherein using analog-modulated that the system transition that data are added on the carrier wave is added to system on the carrier wave for wherein using digital modulation with data.

In radio digital communication system, the standardization air interfaces specify most systems parameter, comprise modulation type, burst format, communication protocol etc.For example ETSI European Telecommunications Standards Institute (BTSI) has stipulated a kind of global system for mobile communications (GSM) standard, it use that Guassian Minimum Shift Keying GMSK (GMSK) modulation scheme character rate uses that time division multiple access inserts that (TDMA) controls on as radio frequency (RF) physical channel of 271ksps or link, the communicating by letter of speech and data message.In the U.S., telecommunications industry association (TIA) has published many interim standards, as IS-54 and IS-136, defined the digital AMPSAdvanced Mobile Phone Service (D-AMPS) of various version, a kind of use difference quadrature phase shift keying (DQPSK) modulation scheme is at the tdma system of RF link.

Tdma system is subdivided into one or more RF channels with available band, and each channel further is divided into the physical channel of a plurality of corresponding tdma frame time slots again.Logic channel has specified the physical channel of modulation and coding to form by one or several.In these systems, travelling carriage is by sending and receive the base station communication of digital information pulse train and a plurality of dispersions on up link and down link RF channel.

At present the growth number of the travelling carriage that uses has produced the demand of multi-voice and data channel more in the cellular telecommunication system.The base station becomes at interval nearer as a result, and making increases in interference adjacent or that operate in the nearer sub-district at interval between the travelling carriage on the same frequency.Although digital technology provides the how useful channel on the given frequency spectrum, also need to reduce and disturb, or more particularly will increase the ratio (being carrier wave-interference (C/I) ratio) of carrier signal strength and interference.

For various communication services are provided, need user's bit rate of corresponding minimum.For example to speech and/or data service, corresponding speech quality of user's bit rate and/or data throughout, user's bit rate is high more, and speech quality is good more and/or data throughout is big more.Total user's bit rate is determined by the combination technique of the speech coding of selecting, error correction coding (chnnel coding), modulation scheme is connected assignable resource with each number (as coding, time slot, frequency etc.).

Concrete care is error correction or chnnel coding in this standard.Digital communication system adopts various technology to handle the wrong information that receives.In general, these technology comprise the technology of the information that those help receivers to correct a mistake to receive, and can make as forward error correction (FEC) technology and those to receive the technology that wrong information resends receiver, as automatic repeat requests technology.The FEC technology comprises for example data convolution or the block encoding before modulation, and these codings are used to allow to correct by noise and disturb the error code that causes.The FEC coding relates to the data bit that the coded-bit that uses greater number is represented some.The common like this encoding rate by them is indicated convolution code as 1/2 and 1/3, and wherein lower code check provides stronger error protection, but then user's bit rate is lower for given channel bit rate.

At receiver side, the sequence that receives is decoded so that allow further deal with data.Certainly the channel of transmission signals makes this decoding processing have more challenge thereon.This challenging diagram is seen Fig. 1.Here, symbols streams u is input to convolution coder 10.Encoded symbols streams x goes up transmission and is received by the form of receiving entity with symbols streams y at a transmission channel 12 (for example an air interface or Wireline), and this symbols streams has reflected the influence of channel to transmission signals.MAP decoder 14 is interpreted as output symbol stream  with the symbols streams y that receives, and this is the estimation of original symbol stream u.Attention is discussed in order to simplify this, has ignored the description of various other processing well known in the art, and for example the modulation and demodulation of carrying out respectively at transmitter and receiver side are handled.

Decoder 14 can use the MAP decoding algorithm, this algorithm is also referred to as by the maximal posterior probability algorithm of symbol or bcjr algorithm, it merges the disclosure content here as a reference by Bahl, Cocke, Jelinek and Raviv proposition in March, 1974, IEEE Trans.Inform.Theory volume IT-20 284-287 page or leaf " minimized the optimal decoding of the uniform enconding of symbol error rate ".Also proposed the various mutation of MAP algorithm, these mutation have more commercial flexibility from the viewpoint of using.For example, merge the disclosure content here as a reference at the Log-Map of in March, 1996 people such as Hagenauer in the works " iterative decoding of binary block and convolution code " of IEEE Trans.Inform.Theory volume IT-42 429-445 page or leaf (carry out handle at log-domain) and Max-Log-Map (simple version of Log-Map).

Because with respect to not too complicated decoder (as the Viterbi decoder), the calculation of complex of MAP decoder and only have quite little performance to increase is so adopted the less relatively of MAP decoder from it since the seventies occurs under most conditions.But along with the continuous development of error correction coding, to the interest of MAP decoder owing to rekindled near the agricultural limit earlier.The professional in present technique field will understand energy per bit and the monolateral noise density (E that the first agricultural limit has determined to be used for the minimum of reliable transmission _b/ N _o) ratio.Although farming has proved this limit at the forties latter stage earlier, also do not develop the error correction coding that this theoretical possibility performance can be provided.But occurred a class what is called " turbo (turbine) " sign indicating number recently, they more approach the operation on the formerly agricultural limit.Because the arrival of these and other novel error correction codings, the MAP decoder has been got back to leading position.

The detailed computing of MAP algorithm is described below, and the formula of reviewing here can be understood measuring of relating in the MAP algorithm better.It is that a decoder of realizing the MAP algorithm in direct mode simply is provided that of MAP decoder may use.That is to say that to the sequence with N symbol that receives, the first step is calculated and stored all branch transition and measures with forward-facing state and measure.Calculated N branch transition measure measure with N forward-facing state after, calculate all N back to status quantitavie.At last, calculate the back behind status quantitavie, calculated likelihood ratio.Even but this direct MAP decoding processing realizes that related processing delay is also quite high when using powerful processor to carry out these calculating.Present communication needs more and more higher bit rate, promptly be need be more and more littler the processing delay tolerance limit.

Reducing of finding in the literature finds in the WO98/20617 disclosure of exercise question for " the soft-decision output decoder of the code word of decoding convolutional encoding " by being intended that of the processing delay related with using the MAP algorithm, and merging is as a reference especially here for its disclosure.Here, described the Log-Map decoder, wherein the viterbi decoder of first " general " is at initial condition t ₀Beginning also provides a plurality of forward-facing states to measure α for each state in each time interval on length of window 2L.The viterbi decoder of second " general " also is provided, and it is at the second time t _2LBeginning also provides a plurality of backs to status quantitavie β for each state in each time interval.Then processor is carried out each state and is asked max calculation twice, uses forward-facing state to measure to first state, to second state use the back to status quantitavie and between to first and second states path use the branch transition state.

Although this MAP realization of decoding realizes reducing processing delay with respect to direct MAP algorithm, its complexity is still higher relatively and say so inferior good after second " general " viterbi decoder uses initial estimation to calculate when beginning at every turn on the meaning of status quantitavie.Therefore be desirable to provide the added complexity with the processing delay that reduces, limited (if there is) and do not need to take the MAP decoder of the inferior good mutation of algorithm to realize making it possible to strong error correction coding is carried out efficient coding, described strong error correction coding is considered for next generation communication system.

Summary

These and other shortcomings and the restriction of the method and system of traditional transmission information can be overcome according to the present invention, processing delay (as with respect to direct applied one approximate 2 factor) can be reduced to decoder and the decoding technique that status quantitavie calculates (RSMC) parallel processing by using a kind of wherein forward-facing state measure calculation (FSMC) and back.Implementation complexity also reduces with respect to for example WO98/20617 and can further reduce time delay by approval processing sequential and the elimination unnecessary extra process unit (as branch transition metric computation unit and likelihood ratio computing unit) related with various tasks.

The accompanying drawing summary

Read following detailed in conjunction with the accompanying drawings, these and other purposes of the present invention, feature and advantage will become more obvious, wherein:

Fig. 1 is that explanation is relating to the general block diagram that sends and receive a signal on the channel that uses error correction coding;

Fig. 2 is the block diagram according to the demonstration decoder of exemplary embodiment of the present invention;

Fig. 3 is the flow chart of describing according to the demonstration decode procedure of exemplary embodiment of the present invention;

Fig. 4 is the sequential chart of timing aspect of operation of the demonstration program of the demonstration decoder of key diagram 2 and Fig. 3;

Fig. 5 is another sequential chart of timing aspect of operation of the demonstration program of the demonstration decoder of key diagram 2 and Fig. 3;

Fig. 6 is the block diagram according to the decoder of another exemplary embodiment of the present invention; And

Fig. 7 is the block diagram that can use exemplary wireless communications systems of the present invention.

Describe in detail

The unrestricted purpose for explanation has in the following description been delivered concrete details such as concrete circuit, circuit element, technology etc. so that thoroughly understand the present invention.But to the professional in present technique field obviously the present invention can realize with other embodiments that are different from detail.In other example, will ignore the detailed description of known method, equipment and circuit for not fuzzy description of the invention.

Should provide some wherein to consider environment more of the present invention so that be given in order to begin this discussion about the additional detail of MAP decoding algorithm and demonstration mutation such as LogMAP algorithm and Max-Log-MAP algorithm.Use some variablees in the following discussion, be defined as follows:

S _k: the state of k node in the grating texture

Measure calculation relates to probability, and its symbol is as follows: A, B and C presentation of events, then to give a definition effectively:

P (A): the probability of incident A

(A, B): incident A and B unite the probability of generation to P

(C): incident A, B and C unite the probability of generation to P for A, B

P (A|B): the conditional probability of incident A when given incident B takes place

(A, B|C): incident A and B united the conditional probability of generation to P when given incident C took place

(A|B, C): given incident B and C unite the conditional probability MAP algorithm that incident A takes place when taking place to P

The MAP algorithm comprises following 4 steps:

1. branch transition measure calculation (BTMC): from the symbol y that receives at channel 12 _kBranch Computed shifts measures γ _k(S _K-1, S _k):

γ _k(S _k-1，S _k)＝P(y _k，S _k|S _k-1)

(1)

＝P(y _k|S _k-1，S _k)·P(S _k|S _k-1)

2. forward-facing state measure calculation (FSMC): initially to measure α ₀(S ₀) the recursive calculation forward-facing state measures α _k(S _k), initially measure initial condition definition according to convolution coder: ${\mathrm{\α}}_k\left(S_k\right)=\underset{S_{k-1}}{\mathrm{\Σ}}{\mathrm{\α}}_{k-1}\left(S_{k-1}\right)\·Y_k(S_{k-1},S_k)---\left(2\right)$

3. (RSMC) calculated to status quantitavie in the back: initially to measure β _N(S _N) after the recursive calculation to status quantitavie β _k(S _k), initially measure end-state definition according to convolution coder: ${\mathrm{\β}}_{k-1}\left(S_{k-1}\right)=\underset{S_k}{\mathrm{\Σ}}{\mathrm{\β}}_k\left(S_k\right)\·Y_k(S_{k-1},S_k)---\left(3\right)$

4. likelihood ratio is calculated (LRC): calculate soft output ∧ _k, the likelihood ratio u of itself and given y _kOf equal value.Note molecule and denominator with respectively at u _k=+1 and u _kThose states of=-1 are to last: ${\mathrm{\Λ}}_k=\frac{P(u_k=+1|y)}{P(u_k=-1|y)}=\frac{\underset{S_{k-1},S_k,u_k=+1}{\mathrm{\Σ}}P(S_{k-1},S_k,y)}{\underset{S_{k-1},S_k,u_k=-1}{\mathrm{\Σ}}P(S_{k-1},S_k,y)}$ (4) $=\frac{\underset{S_{k-1},S_k,u_k=+1}{\mathrm{\Σ}}{\mathrm{\α}}_{k-1}\left(S_{k-1}\right)\·Y_k(S_{k-1},S_k)\·{\mathrm{\β}}_k\left(S_k\right)}{\underset{S_{k-1},S_k,u_k=-1}{\mathrm{\Σ}}{\mathrm{\α}}_{k-1}\left(S_{k-1}\right)\·Y_k(S_{k-1},S_k)\·{\mathrm{\β}}_k\left(S_k\right)}$ The Log-MAP algorithm

If take the logarithm and the symbol a=ln α of all parameters in the formula (1-4), b=ln β, c=ln γ and L=ln Δ, then the formula of BTMC, FSMC, RSMC and LRC becomes 4 steps in following formula (1 '-4 ') and the processing Log-MAP algorithm:

1.BTMC：

c _k(S _k-1，S _k)＝ln(P(y _k|S _k-1，S _k))+ln(P(S _k|S _k-1))(1′)

2. initially measure and be a ₀(S ₀)=ln (α ₀(S ₀)) FSMC: $a_k\left(S_k\right)=\ln \left(\underset{S_{k-n}}{\mathrm{\Σ}}{e}^{a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k)}\right)---\left(2^{\′}\right)$

3. initially measure and be b ₀(S ₀)=ln (β ₀(S ₀)) RSMC: $b_{k-1}\left(S_{k-1}\right)=\ln \left(\underset{S_k}{\mathrm{\Σ}}{e}^{b_k\left(S_k\right)+c_k(S_{k-1},S_k)}\right)---\left(3^{\′}\right)$

4.LRC： $L_k=\ln \left(\underset{S_{k-1},S_k,u_k=+1}{\mathrm{\Σ}}{e}^{a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k)+b_k\left(S_k\right)}\right)-$

(4′) $-\ln \left(\underset{S_{k-1},S_k,u_k=-1}{\mathrm{\Σ}}{e}^{a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k)+b_k\left(S_k\right)}\right)$ The Max-Log-MAP algorithm

If use approximation: $\ln \left(\underset{i}{\mathrm{\Σ}}{e}^{x_i}\right)=\underset{i}{\mathrm{Max}}\left(x_i\right)---\left(5\right)$

That formula (1 '-4 ') becomes is following (1 "-4 "), the Log-MAP algorithm becomes the Max-Log-MAP algorithm. and four steps of then carrying out the Max-Log-MAP algorithm are as follows:

1.BTMC：

c _k(S _k-1，S _k)＝ln(P(y _k|S _k-1，S _k))+ln(P(S _k|S _k-1))(1″)

2. initially measure and be a ₀(S ₀)=ln (α ₀(S ₀)) FSMC: $a_k\left(S_k\right)=\underset{S_{k-1}}{\mathrm{Max}}(a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k))---\left(2^{\′\′}\right)$

3. initially measure and be b ₀(S ₀)=1n (β ₀(S ₀)) RSMC: $b_{k-1}\left(S_{k-1}\right)=\underset{S_k}{\mathrm{Max}}(b_k\left(S_k\right)+c_k(S_{k-1},S_k))---\left(3^{\′\′}\right)$

4.LRC： $L_k=\underset{S_{k-1},S_k,u_k=+1}{\mathrm{Max}}(a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k)+b_k\left(S_k\right))-$

(4″) $-\underset{S_{k-1},S_k,u_k=-1}{\mathrm{Max}}(a_{k-1}\left(S_{k-1}\right)+c_k(S_{k-1},S_k)+b_k\left(S_k\right))$

The same with the many algorithms that find in the document, the MAP theory is interesting, but theoretical realization has proposed realistic problem to the design engineer.In MAP algorithm and above-mentioned mutation thereof, a such problem relates to processing delay.The demonstration of having described relative WO 98/20617 above realizes that this realization wants to reduce the processing delay related with the MAP algorithm.Exemplary embodiment of the present invention provides the another kind of realization that also reduces processing delay, and computation complexity is less and estimated value is not used to status quantitavie calculating in each back.

Figure 2 shows that the block diagram of an exemplary embodiment of the present invention.Because the present invention can be applicable to original MAP algorithm, and Log-MPA and Max-Log-MAP (with other any mutation) algorithm, signal name and reference symbol among the figure that the exemplary embodiment is here provided are described below.To the effective signal name of above-mentioned agreement of MAP algorithm on every signal line with top or with first signal with reference to giving out.After the MAP signal name or under square brackets in provide effective another signal name of the naming convention related with Log-MAP algorithm or Max-Log-MAP algorithm.For example referring to the upper left corner of Fig. 2, the holding wire related with input above BTMC unit 20 has a MAP signal with reference to P (y _k| S _K-1, S _k) and a Log-MAP (or Max-Log-MAP) signal with reference to ln (P (y _k| S _K-1, S _k)).

The demonstration decoder of Fig. 2 has the different functional block of a plurality of following work according to the present invention.BTM is that the mode that the concrete version of the MAP algorithm that uses with decoder in BTMC unit 20 and 22 is determined is calculated.If for example adopt the MAP algorithm, then branch metric calculates according to formula (1), if adopt the Log-MAP algorithm, then according to formula (1 ') if or adopt the Max-Log-MAP algorithm, then calculate according to formula (1 ").

Similarly, calculating forward-facing state in the mode that depends on the MAP mutation that decoder uses in FSMC unit 24 measures.Just, if adopt the MAP algorithm to calculate FSM, if adopt the Log-MAP algorithm to calculate, if use the Max-Log-MAP algorithm to calculate according to formula (2 ") according to formula (2 ') according to formula (2).In RSMC unit 26, calculate the back similarly to status quantitavie, if adopt the MAP algorithm, if adopt the Log-MAP algorithm to calculate, if use the Max-Log-MAP algorithm to calculate according to formula (3 ") according to formula (3 ') according to formula (3).

In LRC unit 28 and 30, use branch metric, the forward-facing state in unit 20,22,24 and 26, determined to measure and reverse state measure calculation likelihood ratio.In particular,, then calculate likelihood ratio,, then calculate,, then calculate according to formula (4 ") if adopt the Max-Log-MAP algorithm according to formula (4 ') if adopt the Log-MAP algorithm according to formula (4) if adopt the MAP algorithm.

Fig. 2 also comprises time delay unit 32 and 34 and memory cell 36 and 38.These unit are used to cushion and control the measurement value of delivering to LRC unit 28 and 30 by timing and control unit 40.For avoiding fuzzy this figure, ignored from the holding wire of other functional blocks of timing and control unit 40 to Fig. 2.But the professional in present technique field will understand the timing (I/O) of each square frame in timing and control unit 40 control charts 2, can calculate and transmit measurement value and regularly and between each functional block of control unit 40 and Fig. 2 provide holding wire in the mode that describes below of for example Fig. 3-5 like this.

In order to understand the working method according to demonstration MAP decoder of the present invention better, Fig. 3 provides the flow chart of an explanation demonstration work decoding.Here receive a sequence deciphering in step 50.This sequence that receives is passable, for example, handles (as down-conversion, demodulation etc.) earlier with the variety of way that the professional in present technique field knows better at receiver.Next in step 52, the initial condition that forward-facing state is measured and the back will be used to the status quantitavie recurrence of unit 24 and 26 is set respectively.Determine the initial condition of FSMC unit from the initial condition (fix typically, so receiver being known) of convolution coder.If end-state is known,, then determine the initial condition of RSMC unit from the end-state of convolution coder as when convolution coder uses the convolution process that stops.Otherwise for example when convolution coder uses hangover convolution process, can use the estimation of end-state.

For example, consider a system, wherein the convolution coder of transmitter has i.e. 4 kinds of different possible the states of 3 constraint length, and wherein receiver adopts the Log-MAP algorithm of the realization according to the present invention.If the initial condition of encoder is for example S ₀=0, then initial FSM can be set to a ₀(S ₀=0)=0 and a ₀(S ₀=1)=a ₀(S ₀=2)=a ₀(S ₀=3)=-∞.If end-state is known as for example S _N=2, then Initial R SM can be set to b _N(S _N=2)=0 and b _N(S _N=0)=b _N(S _N=1)=b ₀(S ₀=3)=-∞.Receiver is not known the end-state of convolution coder else if, and then Initial R SM can be set to b _N(S _N=0)=b _N(S _N=1)=b _N(S _N=2)=b _N(S _N=3)=0.

Get back to Fig. 3 now, as a reference affix k carries out initialization in step 54 pair decode procedure.This affix can for example corresponding time step based on character rate and/or sampling rate selection.Carry out initialization with initial FSM and RSM respectively in this moment FSMC unit 24 and RSMC 26 in step 52 storage.In step 56, increase affix k and decode procedure and begin.The demonstration decode procedure of describing in this embodiment of the present invention is by a periodic duty, and wherein forward-facing state is measured and be back to the status quantitavie parallel computation, has reduced decoding delay like this.Processing method is odd number or even number and slightly different according to symbolic number N in the sequence that receives.

First in circulation promptly comprises step 56-62, and the branch transition of calculating and storing receiving sequence is measured, forward-facing state is measured and a back part to status quantitavie.In particular, (if N is an even number) or k=N/2-1/2 (if N is an odd number) from step-length K=1 to step-length k=N/2 calculate affix and are the RSM (step 60) that the BTM of k and N-k+1 (step 58), FSM (step 60) that affix is k and affix are N-k.FSM and RSM are stored in respectively in unit 36 and 38.In this, as determining that in declaring piece 62 decode procedure has had enough metric data and begun to calculate likelihood ratio, and remaining is measured to be calculated so that process expanded to and comprises that likelihood ratio calculates.

If odd number symbol N is arranged, then only calculate one time likelihood ratio during the iteration in the first time of Extended Cyclic in receiving sequence.In step 64,, use α here like this if k=N/2+1/2 (promptly having only N is that odd number is just for true) handling process forwards step 66 to _K-1(S _K-1), γ _k(S _K-1, S _k) and B _k(S _k) to calculate affix be the LR of N/2+1/2.The flow process circulation turns back to step 56 then increases affix k, calculates and stores one group and newly measure and calculate then extra likelihood ratio.

If on the other hand N be even number or for the first time after the iteration when N is odd number, flow process forwards piece 68 to from piece 64, wherein through two likelihood ratios of the each calculating of circulation.Exactly,, calculate and the storage affix is the BTM of k and N-k+1 up to step-length k=N from step-length k=N/2+1 (if N is an even number) or k=N/2+3/2 (if N is an odd number), affix is that the FSM of k, RSM and the affix that affix is N-k are the LR of K and N-k+1.When decoder arrived step-length k=N, process 70 was come out along "No" branch from Decision Block.Then can be from the sequence after the decoder output decoding in step 72.

As previously mentioned, timing and control unit 40 (Fig. 2) are controlled calculating of measuring and the demonstration decode procedure that the transmission value is described with reference to figure 3 with acquisition between various computing units.Because forward-facing state is measured and the back be that first sign-on from receiving sequence calculates then to status quantitavie is that last sign-on from receiving sequence calculates to the status quantitavie parallel processing and because forward-facing state is measured, therefore for some time at interval, therebetween receiving sequence is calculated and measured, but do not calculate likelihood ratio.As mentioned above, in case affix arrives N/2, just can begin to calculate likelihood ratio.The sequential of exemplary embodiment of the present invention can use the sequential chart of Figure 4 and 5 to see clearlyer, wherein with top Fig. 2 and the same naming convention of 3 uses.

Fig. 4 explanation is when N is even number and measure timing with the likelihood ratio compute associations.Wherein can see at moment k=0, initial forward status quantitavie of FSMC 24 outputs, and RSMC26 therewith exports an initial back to status quantitavie simultaneously.But LRC 28 and 30 does not provide likelihood ratio constantly at this.When affix k increased, FSMC 24 and RSMC 26 continuation output forward directions and back were to measuring.When k=N/2+1, LRC unit 28 and 30 all uses previous forward direction that calculates and store and back to begin to export likelihood ratio to status quantitavie.This will proceed to k=N always.

Fig. 5 explanation is when N is odd number and measure timing with the likelihood ratio compute associations.Wherein can see, again, at moment k=0, initial forward status quantitavie of FSMC 24 outputs, and RSMC 26 therewith exports an initial back to status quantitavie simultaneously.But LRC 28 and 30 does not provide likelihood ratio constantly at this.When affix k increased, FSMC 24 and RSMC 26 continuation output forward directions and back were to measuring.When k=N/2+1, LRC unit 30 first likelihood ratios of output.When k=N/2+3/2, LRC unit 28 and 30 all uses previous forward direction that calculates and store and back to begin to export likelihood ratio to status quantitavie then.This will proceed to k=N always.

Calculating in the likelihood ratio according to aforementioned formula, can see being used for determining that forward-facing state is measured and repeat the back during the calculating section of status quantitavie is calculating likelihood ratio.Exactly, forward-facing state measure with the branch status quantitavie multiply each other (or addition) and the back is common computing (as relatively formula (2) and formula (4)) to multiply each other (or addition) of status quantitavie and branch's status quantitavie.According to another exemplary embodiment of the present invention shown in Figure 6, these common computings only need be carried out once like this, promptly during measure calculation rather than in measure calculation and likelihood ratio computing interval.In this drawing, common reference number is reused in the unit that the decoder with Fig. 2 is had same input and output, and the signal naming convention also is the same.So only FSMC 80 has different reference numbers and different output with RSMC 82.Be further noted that and in the decoder of Fig. 6, ignored time delay unit 32 and 34 that this is because LRC unit 28 and 30 has directly used FSM and RSM respectively.

The professional that other of these exemplary embodiment change the present technique field also is obvious.If for example BTMC unit 20 and 22 processing delay are less than or equal to half of processing delay of FSMC unit and RSMC unit respectively, then can from the decoder Fig. 2 and 6, remove a BTMC unit.For example can provide single BTMC unit to come alternately to provide a branch transition to measure to FSMC unit and RSMC unit.Can provide a multiplexer (not shown) between FSMC unit and RSMC unit, to switch its output in the output of BTMC unit.Certainly, also need need need between the unit that the end Branch Computed of sequence grating texture is measured, switch from unit that the beginning Branch Computed of sequence grating texture is measured and those to the input of single BTMC at those.

Similarly, be less than or equal to half processing delay of the processing delay related, then also can remove one of them LRC unit with FSMC unit and RSMC unit if LRC unit 28 and 30 has respectively.In this embodiment, can provide a demodulation multiplexer (not shown) to come alternately input with the single LRC of being outputted to of FSMC and RSMC unit.According to another exemplary embodiment of the present invention, can use the channelization treatment technology to make up FSMC and RSMC unit further to reduce complexity according to decoder of the present invention.

Decoding technique according to the present invention may be used on any communication system and/or environment.But as mentioned above, these MAP decoding techniques (with its mutation) can find concrete application in wireless communications environment, transmission channel 12 is that an air interface and encoder 10 and decoder 14 are parts (or vice versa) of base station and travelling carriage in this environment.For some versatilities are provided, Fig. 7 and following description provide some can realize the general discussion of a kind of like this demonstration system of the present invention.

Fig. 7 represents a block diagram that comprises the demonstration cellular mobile radiotelephone system of an exemplary base station 110 and travelling carriage 120.The base station comprises control and processing unit 130, and it is connected to MSC 140, the latter and then be connected to the PSTN (not shown) again.The general features of this cellular radiotelephone system is being known in the art, as authorize the United States Patent (USP) 5th of people's such as Wejke exercise question for " the adjacent auxiliary switching in the cellular communication system ", 175, No. 867 and authorize the United States Patent (USP) 5th of people's such as Raith exercise question for " a kind of method that in wireless communication system, communicates ", 603, describe in 081, the disclosure of these two patents merges in this application as a reference.

A plurality of voice/datas (promptly professional) channel is handled by a Traffic Channel transceiver 150 in base station 110, and transceiver is by control and processing unit 130 controls.Also have each base station to comprise a control channel transceiver 160, it can handle more than one control channel.Control channel transceiver 160 is by control and processing unit 130 controls.Control channel transceiver 160 on the control channel of base station or sub-district to the travelling carriage broadcast control information that locks onto this control channel.Should be appreciated that transceiver 150 and 160 can realize with individual equipment, as professional and control transceiver 170, and transceiver needs not be special use to the distribution of antenna.As the part of the signal processing of wherein carrying out, base station 110 can comprise a decoder, as mentioned above, is used to remove the error correction coding of the signal correction connection that transmits with travelling carriage 120 on control channel or Traffic Channel.

Travelling carriage 120 is received in the information of broadcasting on the control channel in its business and control channel transceiver 170.Again, receive the part of information processing as this, travelling carriage 120 can comprise the decoder that above-mentioned relatively any previous exemplary embodiment is described.Then, the control channel information (this information comprises the feature of the candidate cell that travelling carriage will lock) that processing unit 180 assessments receive, and which sub-district definite travelling carriage should be locked in.In case distributed a Traffic Channel, the information that processing unit 180 is also deciphered and demodulation receives on Traffic Channel for travelling carriage 120.

Although only describe the present invention in detail with reference to several exemplary embodiment, the professional in the present technique field will understand not departing under the situation of the present invention can carry out various modifications.For example above-mentioned processing can be used to decipher the received signal sequence through turbine coding, wherein will calculate likelihood ratio for a sequence before the value of symbol that receives is carried out any hard decision.So as can be seen the present invention can be applied to equally wherein with hard symbol judgement in series, before carrying out hard symbol judgement or even do not carry out hard decision and produce the decoding technique of soft information.Therefore the present invention only wants that by following the claim that comprises all its equivalents is defined.

Claims (25)

1. the method for the symbol sebolic addressing that receives of a decoding comprises step:

Calculating first and second branch transition measures;

Measuring definite forward-facing state based on described first branch transition measures;

With determine step that described forward-facing state is measured concurrently, measure based on described second branch transition and to determine that the back is to status quantitavie;

Based on described branch transition measure, described forward-facing state is measured and described back to the combination calculation likelihood ratio of status quantitavie;

Use described likelihood ratio to decipher the symbol sebolic addressing of described reception.

2. the process of claim 1 wherein that described branch transition is measured, described forward-facing state is measured, described back to status quantitavie and described likelihood ratio according to the MAP algorithm computation.

3. the process of claim 1 wherein that described branch transition is measured, described forward-facing state is measured, described back to status quantitavie and described likelihood ratio according to the Log-MAP algorithm computation.

4. the process of claim 1 wherein that described branch transition is measured, described forward-facing state is measured, described back to status quantitavie and described likelihood ratio according to the Max-Log-MAP algorithm computation.

5. the process of claim 1 wherein that the step that described first and second branch transition of described calculating are measured also comprises step:

Described first and second branch transition of parallel computation are measured.

6. the process of claim 1 wherein that the step that described first and second branch transition of described calculating are measured also comprises step:

Alternately calculating described first and second branch transition measures.

7. the process of claim 1 wherein that the step of the described likelihood ratio of described calculating also comprises step:

Two likelihood ratios of parallel computation.

8. the process of claim 1 wherein that the step of the described likelihood ratio of described calculating also comprises step:

Alternately calculate:

Based on first likelihood ratio that described forward-facing state one of is measured, described back one of is measured to one of status quantitavie and described first branch transition; And

Based on second likelihood ratio that described forward-facing state one of is measured, described back one of is measured to one of status quantitavie and described second branch transition.

9. the process of claim 1 wherein and describedly determine that described forward direction and back use the channelization treatment technology to carry out to the parallel step of status quantitavie.

10. the method for claim 1 also comprises step:

Storing described forward-facing state measures with described back to status quantitavie; And

After measuring and be described, behind status quantitavie, calculates the described forward-facing state of having stored predetermined number described likelihood ratio.

11. the method for claim 10 also comprises step:

Storing described first and second branch transition measures.

12. the process of claim 1 wherein that described decoding step also comprises step:

The value of symbol of the symbol sebolic addressing that hard decision receives after having calculated all likelihood ratios of described sequence.

13. decipher N method that receives the sequence of symbol and comprise step for one kind:

(a) initialization affix k;

(b) increase described affix k;

(c) calculating first branch transition based on the sequence of described reception measures;

(d) calculating second branch transition based on the sequence of described reception measures;

(e) measuring definite forward-facing state based on described first branch transition measures;

(f) the described forward-facing state of storage is measured;

(g) with determine step that described forward-facing state is measured concurrently, measure based on described second branch transition and to determine that the back is to status quantitavie;

(h) the described back of storage is to status quantitavie;

(i) repeating step (b)-(h) is up to k≤N/2, measures with described back based on described forward-facing state then to begin to calculate likelihood ratio to status quantitavie; And

(j) use described likelihood ratio to decipher the symbol sebolic addressing of described reception.

14. the method for claim 13, wherein said branch transition is measured, described forward-facing state measure described back to status quantitavie and described likelihood ratio according to the MAP algorithm computation.

15. the method for claim 13, wherein said branch transition is measured, described forward-facing state is measured with described likelihood ratio according to the Log-MAP algorithm computation.

16. the method for claim 13, wherein said branch transition is measured, described forward-facing state measure described back to status quantitavie and described likelihood ratio according to the Max-Log-MAP algorithm computation.

17. the method for claim 13, the step that described first and second branch transition of wherein said calculating are measured also comprises step:

Described first and second branch transition of parallel computation are measured.

18. the method for claim 13, the step that described first and second branch transition of wherein said calculating are measured also comprises step:

Using same processing unit alternately to calculate described first and second branch transition measures.

19. the method for claim 13, the step of the described likelihood ratio of wherein said calculating also comprises step:

Two likelihood ratios of parallel computation.

20. the method for claim 13, the step of the described likelihood ratio of wherein said calculating also comprises step:

Alternately calculate:

Based on first likelihood ratio that described forward-facing state one of is measured, described back one of is measured to one of status quantitavie and described first branch transition; And

Based on second likelihood ratio that described forward-facing state one of is measured, described back one of is measured to one of status quantitavie and described second branch transition.

21. the method for claim 13, wherein said definite described forward direction and back use the channelization treatment technology to carry out to the parallel step of status quantitavie.

22. the method for claim 13, wherein said decoding step also comprises step:

The value of symbol of the symbol sebolic addressing that hard decision receives after having calculated all likelihood ratios of described sequence.

23. the transceiver of a received signal comprises:

The reception processing unit of the sequence that reception will be deciphered; With

A MAP error correcting deocder from back to status quantitavie that decipher described sequence by parallel computation forward direction and.

24. the transceiver of claim 23, wherein said MAP error correcting deocder use described forward direction of Log-MAP algorithm computation and back to status quantitavie.

25. the transceiver of claim 23, wherein said MAP error correcting deocder use described forward direction of Max-Log-MAP algorithm computation and back to status quantitavie.

Images